基于标记提取分水岭算法的医学图像分割

针对医学图像分割中分水岭算法的过分割问题，提出了一种改进的基于标记提取的分水岭算法．此算法埘分水岭算法进行厂两点改进：一是在预处理阶段，先对原始图像做腐蚀滤波和膨胀滤波处理，将原始吲像与腐蚀图像相加后减去膨胀图像得到合并图像，再对合并图像做开、闭运算，从而达到滤除原始图像中的噪声和非感知信息、保留原始图像结构信息的目的；二是在分割阶段，对开、闭运算后的图像进行分水岭分割，该分割由形态梯度计算、标记提取和分水岭分割二部分组成．改进后的算法因为无需进行分割后的区域合并处理，降低了分割的复杂性，同时能较好地抑制慨部医学图像中的过分割问题，并把图像中的病变区域有效分割出来．

Research of Medical Image Segmentation Based on Marker Extraction Watershed Algorithm

In view of the medical image segmentation in the over-segmentation of the watershed algorithm, this paper proposes an improved watershed algorithm based on marker extraction. Improved algorithm of watershed algorithm are two improvements： First, the preprocessing stage： the original image is handled by corrosion filter and expansion filter. Then, the original image adds the corrosion image and subtracts expansion image. The result is called Merging image, that is handled by open operation and close operation so as to eliminate the noise and the perceptual information in the original image, to retain the original structural information of the image; Second, segmentation stage： After the opening and closing operation, the image is segmented by Watershed Algorithm, that is composed of three parts.. morphological gradient calculation, Marker extraction and watershed segmentation. Improved algorithm doesn＇t need merge segmentation region, therefore, reducing the complexity of segmentation. At the same time, the improved algorithm can efficiently suppress over-segmentation phenomenon in a medical image of the brain, and makes the lesion area of the image to be effectively split.